Modeling, Analysis and Design for Carrier Aggregation in Heterogeneous Cellular Networks

TL;DR

This paper develops a load-aware model for carrier aggregation in heterogeneous cellular networks, revealing insights into biasing effects, deployment configurations, and strategies to optimize user data rates and network capacity.

Contribution

It introduces a novel load-aware model for multi-band HetNets, analyzing the impact of biasing, deployment strategies, and spectrum allocation on network performance.

Findings

Biasing can increase small cell spectral efficiency with large biasing.

Peak data rate is independent of base station density and transmit power.

Spatial reuse with small cells can outperform spectrum addition in high small cell density scenarios.

Abstract

Carrier aggregation (CA) and small cells are two distinct features of next-generation cellular networks. Cellular networks with small cells take on a very heterogeneous characteristic, and are often referred to as HetNets. In this paper, we introduce a load-aware model for CA-enabled \textit{multi}-band HetNets. Under this model, the impact of biasing can be more appropriately characterized; for example, it is observed that with large enough biasing, the spectral efficiency of small cells may increase while its counterpart in a fully-loaded model always decreases. Further, our analysis reveals that the peak data rate does not depend on the base station density and transmit powers; this strongly motivates other approaches e.g. CA to increase the peak data rate. Last but not least, different band deployment configurations are studied and compared. We find that with large enough small cell density, spatial reuse with small cells outperforms adding more spectrum for increasing user rate. More generally, universal cochannel deployment typically yields the largest rate; and thus a capacity loss exists in orthogonal deployment. This performance gap can be reduced by appropriately tuning the HetNet coverage distribution (e.g. by optimizing biasing factors).